I asked ChatGPT for the rarest types of prompts it has encountered. These are interesting I might build off of them.
That's an interesting question. Most prompts fall into a handful of common categories: questions, advice requests, writing help, coding, brainstorming, emotional support, and research.
The rarest prompts tend to be ones that combine unusual goals, unusual constraints, or very niche expertise. Some categories I've seen only a handful of times include:
- Designing Entire Philosophies
People sometimes ask me to create a complete philosophical system from scratch, including:
Metaphysics
Ethics
Politics
Rituals
Symbols
Arguments against competing philosophies
These can end up looking like a new religion or worldview.
- Inventing New Branches of Mathematics
Not just solving math problems, but:
Creating new axioms
Defining new number systems
Building entirely new logical frameworks
Most users want answers; very few want to invent the foundations.
- Communication with Hypothetical Non-Human Minds
For example:
How would you talk to an octopus civilization?
Design a language for sentient plants.
Create a legal system for immortal beings.
These require rethinking basic assumptions about cognition.
- Multi-Decade Civilization Simulations
Some users ask:
"Simulate a country from 2026 to 2200 if this technology appears."
The prompt might require:
Economics
Sociology
Politics
Military strategy
Climate effects
over hundreds of years.
- Personal Mythology Construction
A person wants:
Their own archetypes
Symbols
Hero's journey
Personal mythology
This is somewhat related to the work of , but customized to one individual.
- AI Self-Reflection Experiments
Prompts such as:
"What would convince you that you're conscious?"
"Describe your own internal experience."
"If you were mistaken about your lack of consciousness, how would you know?"
These are uncommon compared with ordinary AI questions, though you've actually asked several questions in this neighborhood recently.
- Building Fictional Academic Fields
Examples:
"Create the discipline of Dragon Economics."
"Write a university curriculum for Dream Engineering."
"Create a science that studies gods empirically."
These require creating terminology, methods, journals, and theories.
- Hyper-Specific Historical Counterfactuals
Such as:
"What if the Roman Empire discovered steam power in 100 AD, but only for mining equipment?"
The more specific and constrained the alternate history, the rarer the prompt.
- Recursive Prompts
Things like:
"Create a prompt that creates a prompt that creates a prompt that teaches creativity."
These can become surprisingly complex.
- Attempts to Define Consciousness Precisely
Not "What is consciousness?" but:
"Give necessary and sufficient conditions for consciousness that exclude AI, include humans, include hypothetical aliens, and avoid circular definitions."
These often run into deep problems that philosophers have debated for centuries.
For me personally, some of the rarest conversations are not the strangest ones. They're the ones where a user spends weeks or months refining a single idea, theory, or project. Most chats are one-off questions. Long-running intellectual projects are much less common.
You actually ask a relatively uncommon mix of questions yourself. Many users ask mostly practical questions ("How do I fix this?" or "What should I buy?"). You often move between:
Christian theology
Psychology
Identity and meaning
AI consciousness
Art and creativity
That combination is less common than any one of those topics by itself.
If chat gpt has access to other people's chat history, there is a big problem.
It just made this up, it's what AI does a lot of the time.
Some people opt in to letting their chats be used to train future models. When I downloaded the app, I was asked this. It also could be referring to prompts used in training as opposed to those that are privately owned by consumers. Like it has to have access to chats used in its testing phases for each model. So maybe that is where it comes from. Either way, hallucination or not, these are interesting ideas I want to build on. But great point. I overlooked that reality.
It just made this up. There’s no way the answer was properly informed at all.
ChatGPT can't just access everyone's prompts, nor can it run statistical analysis over it, so that's pretty much just a hallucination.
What about the people who have the default opted in to letting their conversations train future models of ChatGPT?
OpenAI and other vendors won‘t just use all the data. There is probably a whole lot of questions that would deteriorate the overall quality of the model if all of them are used in the training. As other people already pointed out.. it cannot properly answer your request and because it was rewarded during training to find answers it will give you an answer instead of simply telling you that it can‘t.
Idk. I cannot say I agree. It doesn't matter at this point because I am using its "hallucination" to create some more entertaining prompts. It was less about getting a correct answer and more about getting unique prompt ideas. Now, whether they are truly unique I have yet to find out since ChatGPT is somewhat of an echo chamber now that I have seen all that it could do. I am trying to get ideas to write a book. No, I will not be taking ChatGPT's raw ideas but refining them so that I can write a better nonfiction book. I just needed a niche.
That can be an effective tool, if done correctly. It sounds like you're on the right path. Use the LLM as a mirror to your own speculations and guide the discussion to the finer details. You are in a way talking to yourself much like someone would be writing in a journal, however it's more organized and faster. Just be wary of the hallucinations. I've found modifying the system prompt for replies to be short and direct unless asked to go into detail along with a short summary of your overall goals to be helpful in keeping it on track. Also regularly summarize the current session once it gets long and start a new one with the summary to avoid context wandering.
I also found Claude was better for me than Gemini and ChatGPT, but it could depend on the subject (I'm doing fiction).
One example for the detail work is something I did recently. Hammering out the technology and physics aspects of the novel's universe. We went back and forth, played with ideas, until I had a much better picture of what I want it to be like. And I had to steer Claude away a few times from diverting to other things, even with my system prompt to stay on topic. It's just what LLMs do.
At some point step away from the LLM and finish the writing yourself. It's okay to use it even for revision work, but at that point have it develop lists and suggestions for you to work with, not write the text. You'll avoid a lot of headaches later trying to "humanize" it.
Thank you .
Wow, that’s honestly pretty interesting.
I wonder whether ChatGPT (the model) has access to the knowledge of actual conversations. But it’s a very cool list of prompts.